get SSL Broken pipe error when try to make push notification - iphone

We develop an iPhone app, and have push notification for development and ad hoc version working properly. But when we try to send push notification to real user devices in our database, we got SSL connection reset, then Broken pipe error. We think maybe there are too many devices in our database (more than 70000), so it is failed to send all messages at the same time. So we try to send messages to 1000 devices once, but still got this "Broken pipe" error for around 100 messages. And we are not sure whether the messages have been send. Any suggestion?

We have solved this problem. It is broken due to invalid tokens in our database table.And apple apns service will disconnect us if there is an invalid token. Since the connection is broken, it will have "Broken Pipe" error when you try to send messages again. Basic solution is to find whether writing to pipe is successful, if not, just disconnect and reconnect again after some delay.

emagic is correct, one reason for the "broken pipe" error can occur from invalid tokens. There are a few other reasons it can occur too. The following is from Apple Technical Note TN2265:
The most common problem is an invalid device token. If the token came
from the sandbox environment, such as when you are testing a
development build in house, you can't send it to the production push
service. Each push environment will issue a different token for the
same device or computer. If you do send a device token to the wrong
environment, the push service will see that as an invalid token and
discard the notification.
An invalid device token can also mean that the user has deleted your
app from their device or computer. You should check the feedback
service at least once a day for device tokens that are no longer
valid.
Other possible issues might be sending a payload longer than 256
bytes, your payload might not be formatted correctly, or perhaps your
JSON dictionary has incorrect syntax.
An occasional disconnect while your provider is idle is nothing to be
concerned about; just re-establish the connection and carry on. If one
of the push servers is down, the load balancing mechanism will
transparently direct your new connection to another server assuming
you connect by hostname and not by static IP address.

Extending on emagic's answer, this is my php code snipplet:
private $fp;
private function connect(){
$apnsHost = 'gateway.push.apple.com';
$apnsCert = 'certs/cert.pem';
$apnsPort = 2195;
$pass = "blah";
$streamContext = stream_context_create();
stream_context_set_option($streamContext, 'ssl', 'local_cert', $apnsCert);
stream_context_set_option($streamContext, 'ssl', 'passphrase', $pass);
$this->fp = stream_socket_client('ssl://' . $apnsHost . ':' . $apnsPort, $err, $errstr, 60, STREAM_CLIENT_CONNECT|STREAM_CLIENT_PERSISTENT, $streamContext);
if (!$this->fp) return("Failed to connect: $err $errstr<br>");
echo 'Connected to APNS<hr />';
}
private function send(){
$this->connect();
foreach($pushes as $push) {
$payload['aps'] = array('alert' => $push->text, 'badge' => 0, 'sound' => 'default');
$payload2 = json_encode($payload);
$msg = chr(0) . pack('n', 32) . pack('H*', $push->token) . pack('n', strlen($payload2)) . $payload2;
try {
$result = fwrite($this->fp, $msg, strlen($msg));
}
catch (\Exception $e) {
fclose($this->fp);
echo('Error sending payload: ' . $e->getMessage());
sleep(5);
$this->connect();
}
}
}

Related

Why is IO::Socket::Socks unreliable when creating a proxy chain?

I want to use this module to create a connection using multiple SOCKS proxies, but the module I am using is very unreliable and slow compared to using an external solution such as proxychains. After successfully connecting through the chain of proxies, the while loop begins sending the data through the proxy chain. However, the program crashes with the error: "Out of memory". Does anyone have advice on how to fix the code? Any help would be appreciated.
while ($sock) {
$sock->syswrite (
"GET / HTTP/1.1\r\n"
);
}
When I look at the Wireshark capture, the initial request is sent without any problems but following requests have duplicate http requests in each packet. See: Wireshark capture
Edit: I tried using IO::Async, but the requests aren't sent.
my $loop = IO::Async::Loop->new;
my $handle = IO::Async::Handle->new(
write_handle => $sock,
on_write_ready => sub {
my $request = "GET / HTTP/1.1\r\n"
print $sock $request;
}
);
$loop->add( $handle );
$loop->loop_forever;
Edit: I was able to fix the script by establishing a new chain for each request I send. Is it possible to send a CONNECT request through the socks chain to establish a new connection each time I send a http request?

Perl Net::SSLeay check if socket is available to read

I am using a perl module called Net::APNS::Persistent. It helps me to open up a persistent connection with apple's apns server and send push notifications through APNS. This module uses Net::SSLeay for ssl communication with APNS server.
Now, I want to read from my socket periodically to check if APNS sends back any response. Net::APNS::Persistent already has a function called _read() which looks like below:
sub _read {
my $self = shift;
my ($socket, $ctx, $ssl) = #{$self->_connection};
my $data = Net::SSLeay::ssl_read_all( $ssl );
die_if_ssl_error("error reading from ssl connection: $!");
return $data;
}
However, this function works only after APNS drops the connection and I get error while trying to write. On other times my script gets stuck at,
my $data = Net::SSLeay::ssl_read_all( $ssl );
I checked Net::SSLeay doc and found it has a method called peek
Copies $max bytes from the specified $ssl into the returned value. In contrast to the Net::SSLeay::read() function, the data in the SSL buffer is unmodified after the SSL_peek() operation.
I though it might be useful, so I added another function within the Net::APNS::Persistent module:
sub ssl_peek {
my $self = shift;
my ($socket, $ctx, $ssl) = #{$self->_connection};
print "Peeking \n";
my $data = Net::SSLeay::peek( $ssl, $pending );
print "Done peeking \n";
return $data;
}
Unfortunately this also gave me the same problem. It only prints Peeking and never reaches the line where it would print Done peeking. Had same problem using Net::SSLeay::read. Is there a way to check if the socket can be read or maybe set a read timeout so that my script doesnt get stuck while trying to read from socket?
The APNS documentation says the following:
If you send a notification that is accepted by APNs, nothing is returned.
If you send a notification that is malformed or otherwise unintelligible, APNs returns an error-response packet and closes the connection. Any notifications that you sent after the malformed notification using the same connection are discarded, and must be resent
As long as your notifications as accepted, there won't be any data to read and thus a read operation on the socket will block. The only time there's data available is when there's an error, and then the connection is immediately closed. That should explain the behaviour you're observing.
To check if the underlying socket can be read use select, i.e.
IO::Select->new(fileno($socket))->can_read(timeout);
timeout can be 0 to just check and not wait, can be a number of seconds or can be undef to wait forever. But before you do the select check if data are still available in the SSL buffer:
if (Net::SSLeay::pending($ssl)) { ... use SSL_peek or SSL_read ... }
Apart from that it does look like that the module you use does not even attempt to validate the servers certificate :(

Python APNs does not process request

I'm trying to implement a server side script for sending push notifications to apple push notification server. I create the ssl connection, I send the payload - but am unable to get a response from the APNs. Here is my code:
import socket, ssl, pprint, struct, time, binascii
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
# require a certificate from the server
ssl_sock = ssl.wrap_socket( s,
keyfile="/Users/Jeff/Desktop/pickmeup-key2-noenc.pem",
certfile="/Users/Jeff/Desktop/pickmeup-cert2.pem",
server_side=False,
do_handshake_on_connect=True,
cert_reqs=ssl.CERT_REQUIRED,
ca_certs="/Users/Jeff/Desktop/entrustrootcert.pem",)
#ciphers="ALL")
ssl_sock.connect(('gateway.sandbox.push.apple.com', 2195))
print repr(ssl_sock.getpeername())
print ssl_sock.cipher()
print pprint.pformat(ssl_sock.getpeercert())
command = '\x00'
identifier = 1987
expiry = time.time()
deviceToken = "9858d81caa236a86cc67d01e1a07ba1df0982178dd7c95aae115d033b93cb3f5"
alert = "This is a test message"
sound = "UILocalNotificationDefaultSoundName"
payload = "{\"aps\":{\"alert\":\"%s\",\"sound\":\"%s\"}}" %(alert, sound)
packetFormat = "!cIIH%dsH%ds" %(32, len(payload))
packet = struct.pack(packetFormat,
command,
identifier,
int(expiry),
32,
binascii.unhexlify(deviceToken),
len(payload),
payload)
nBytesWritten = ssl_sock.write(packet)
print "nBytesWritten = %d" %(nBytesWritten)
data = ssl_sock.read(1024)
print len(data)
ssl_sock.close()
Running this script, I generate the following output:
('17.149.34.132', 2195)
('AES256-SHA', 'TLSv1/SSLv3', 256)
{'notAfter': 'May 31 00:04:27 2012 GMT',
'subject': ((('countryName', u'US'),),
(('stateOrProvinceName', u'California'),),
(('localityName', u'Cupertino'),),
(('organizationName', u'Apple Inc'),),
(('organizationalUnitName', u'Internet Services'),),
(('commonName', u'gateway.sandbox.push.apple.com'),))}
nBytesWritten = 133
0
Any ideas on what might be going wrong? (I am sending enhanced push notifications so I am expecting a response from apple push notification server)
The key thing to note is that read() is returning no data. In Python, read() is supposed to block until data is available or the connection closes. Apple is closing your connection.
Why? Well, probably because you sent a malformed request. command=0 is a normal push notification; command=1 is enhanced. The big-endian 1987 will be interpreted as a 0-byte device token and a 1987-byte payload, neither of which are valid.
(And FWIW, I'd use B instead of c for the command ID; it seems to make more sense.)
you may consider https://github.com/djacobs/PyAPNs that wrapped lot of useful features, including:
error handling
support enhanced message format and auto resend messages which are sent before error response
non-blocking ssl socket connection with great performance
Apple Push notification server doesn't give a response, it's a one-way binary socket.
Rather than rolling your own solution you could try apns-python-wrapper or apns

How do I check if an email address is valid without sending anything to it?

I have a client with 5000 emails from an old list he has that he wants to promote his services to. He wants to know which emails on the list are still valid. I want to check them for him - without sending out 5K emails randomly and then being listed as a spammer or something. Ideas?
You can validate the email via SMTP without sending an actual email.
http://code.google.com/p/php-smtp-email-validation/
You could also send emails out, and check for bounces.
bucabay's answer is the way forward. What a library like that essentially does is checking for existing DNS record for (mail) servers at specified domains (A, MX, or AAAA). After that, it do what's termed callback verification. That's where you connect to the mail server, tell it you want to send to a particular email address and see if they say OK.
For callback verification, you should note greylisting servers say OK to everything so there is no 100% guarantee possible without actually sending the emails out. Here's some code I used when I did this manually. It's a patch onto the email address parser from here.
#
# Email callback verification
# Based on http://uk2.php.net/manual/en/function.getmxrr.php
#
if (strlen($bits['domain-literal'])){
$records = array($bits['domain-literal']);
}elseif (!getmxrr($bits['domain'], $mx_records, $mx_weight)){
$records = array($bits['domain']);
}else{
$mxs = array();
for ($i = 0; $i < count($mx_records); $i++){
$mxs[$mx_records[$i]] = $mx_weight[$i];
}
asort($mxs);
$records = array_keys($mxs);
}
$user_okay = false;
for ($j = 0; $j < count($records) && !$user_okay; $j++){
$fp = #fsockopen($records[$j], 25, $errno, $errstr, 2);
if($fp){
$ms_resp = "";
$ms_resp .= send_command($fp, "HELO ******.com");
$ms_resp .= send_command($fp, "MAIL FROM:<>");
$rcpt_text = send_command($fp, "RCPT TO:<" . $email . ">");
$ms_resp .= $rcpt_text;
$ms_code = intval(substr($rcpt_text, 0, 3));
if ($ms_code == 250 || $ms_code == 451){ // Accept all user account on greylisting server
$user_okay = true;
}
$ms_resp .= send_command($fp, "QUIT");
fclose($fp);
}
}
return $user_okay ? 1 : 0;
I think you need to send the emails to find out. Also, this is pretty much exactly what a spammer is, thus the reason for getting put on spammer lists. Sending in bursts will help you hide this fact though.
You'll have to email them at least once.
Create a new email list. Send the old list an email with a link they need to click on to continue receiving messages (re-subscribe).
Send them all an email and collect all reply-to bounces on a real email account, then purge those bounced emails from your main list.
Send them all an HTML email, and one of the images is remotely hosted and requires a unique ID to request it that you set in each email. When your web server returns that image to their client, you can then consider that email as active. This is called a web bug, and will only work if the person automatically loads remote images in their client.
https://github.com/kamilc/email_verifier is a rubygem that will check that the MX record exists and that the SMTP server says the address has a valid mailbox.
You can use a paid service like Kickbox to do this as well.
You can consider the MailboxValidator service http://www.mailboxvalidator.com/ which should be adequate for your requirement. You can get either a bulk plan where you can upload a CSV file containing your email list or get the API plan if you require programmatic integrations.

Using CakePHP's Email component

I try to send a simple Email via CakePHP's Email Component. I'm using following code from the cookbook documentation:
$this->Email->from = 'Irgendjemand <irgendjemand#example.com>';
$this->Email->to = 'Irgendjemand Anderes <irgendjemand.anderes#example.com>';
$this->Email->subject = 'Test';
$this->Email->send('Dies ist der Nachrichtenrumpf!');
The send()-method does only return a boolean value with the value false - but no error or warning occurs.
Does somebody have a solution for that?
Have you tried changing the delivery options? There are three options: mail, smtp and debug.
$this->Email->delivery = 'debug';
$this->Email->send('test message');
debug($this->Session->read('Message.email'));
You can debug with EMail. Set the delivery to debug and the email message will be set to Session.message:
if (Configure::read('debug') > 1) {
$this->Email->delivery = 'debug';
}
$ret = $this->Email->send();
if (Configure::read('debug') > 1) {
pr($this->Session->read('Message.email'));
}
Which OS are you on? If Windows, this note may be of interest:
Note: The Windows implementation of mail() differs in many ways from the Unix implementation.
...
As such, the to parameter should not be an address in the form of
"Something <someone#example.com>". The mail command may not parse this properly while talking with the MTA.
Secondly, it may just be the case that no mail server will accept outgoing mail from your local machine due to spam protection. I have often seen that the same mail() function will not work locally, but works fine once uploaded to a trustworthy server. You could try to use an authenticated mail relay in that case (SMTP).